As a Japanese I'd say he would be "Naruto Engineer". I find it always awesome that Tony takes a lot of effort, patience and makes it out✨ instead of giving it up as "not fixable". Someone may misunderstand: He's not meaning AMD card is too bad to use. It's just more complicated and paintaking to diagnose an error and fix it, which is why the title says i dont like "fixing" AMD cards.
@@Kosphy There is one kid that licks AMD boots. Then there is the OP... Displaying a strong cognitive bias. Overlooking the repair is need, then overlooking all the missing data points the competition provides. A bias so strong only a direct statement against the manufacture would be considered.... because all the other evidence does not exist.
The thing why that controller fail a lot is because it's designed to be used with ir/infineon drmos/sps and not sic649 or sic632 so they can comunicate better. Also in general for me repairing AMD cards are easier than nvidia simply because most nvidia cards i saw are dead cores instead of AMD most of them are just dying transistor or shorted drmos.
It's also the same on nvidia cards btw, MPS controller are designed to be together with MPS SPS/drmos so when they fail their protection can kick in properly
@@accurian148I have seen more MP86990 DrMos (coupled with an MPS2888A) burn a hole through a card (mostly zotac cards) than other brands of Mosfets/controllers, causing several layers to short together and making the repair almost impossible. (Good luck reconfiguring an MPS controller to run with 1 missing phase).
@@gpudoctor doesn't shock me at all tbh, Zotac use cheap components and design their cards with the same philosophy unfortunately. It wouldn't be so bad but they also overclock the cards aggressively making them ticking time bombs, especially for the higher end cards.
I respect showing both the soldering failures and even things that may seem silly like forgetting to solder the dongle back in. Incredibly informative and great walk through!
GPU = Initialism (abbreviation) of "graphics processing unit". Plural: Two GPUs (which will be written out: ""graphics processing units") Easy, or?:) For reference see Wiktionary.
39:17 Your first flash with the original chip failed as well, but you didn't catch it. I suspect something in the software or hardware of your flashing setup, not the target chips themselves.
Very informative video. While it's more complicated to diagnose, I don't think it matters much for the end user. Also 52°C and 70°C hotspot? Those are pretty good temps right there.
Doubling fases is common practice, I'm not a board tech, but I immediatly spotted the fact they called it 5+1 and there's actually 14, mosfets, inductors and caps. However, like you say, the 2 SOC phases arent in the 5+1 equation. The +1 part is the memory VRM, which exists out of 2 mosfets as well. And the remaining 5 VRM phases are doubled and divided over 10 mosfets.
A lot of 6000 series cards made during the pandemic, mainly the original 6700/6800/6900 variants, suffered a lot from subpar quality components being used because less people were working In the factories, and there was much less oversight to ensure quality control. I've seen many of these cards fail due to many issues including bad silicon, faulty bios chips, and just subpar quality builds Even with brands like Sapphire, PowerColor, XFX, and ASRock which supply the majority of AMD's GPUs due to these companies being strictly AMD only. I would actually suggest to anyone owning any card from the 6000 series to look for the 6x50 variant of any card because the build quality will be much better.
@@gabrielgabriel9779 if AsRock isn't quality i dont know what is. Especially when it comes to boards they use high quality components where it matters as caps and mosfets while keeping a reasonable price for the customer.
@@gabrielgabriel9779 I got a ASRock Challenger Pro D RX 6600 XT when it came out and it is one of the best ones around. Used it for gaming and mining. VRAM temps never came above 70C and gpu usually stayed around 50/55C fans hardly running. ASRock ever since they entered the gpu market have produced good solid cards. The RX 6600 XT now runs in my sons rig without any issues. Times have changed Asus has gone down and ASRock is now with the premium brands.
I don't know if it is just reflections but the cap to the left of phase 3 you identified looks to be bulged. It just looks a little different than the other caps, might be it is not sitting fully flat compared to the rest so reflecting the light differently. Maybe it does look a little domed at 34:29 when the light moves across it.
@@Ren-kei It's like "Sacon FZ cap fiasco" of the GeForce 7-series days all over again! (and for ATI at that time, even when AMD bought them already in 2006, I guess Radeon X1000 series? Or very-early Radeon HD series? Such as Radeon HD 2000 series?)
Рік тому+15
uurgghh these AMD card are the bane of my existence, RX 5000 series upward are a joke to fix, even with the programmer it's still a pain
When flashing at 39:18 it said “Device 1 User section programming failed” followed by many more errors. Also isn’t it unwise to dump and then reuse a potentially corrupted firmware? Of course I have no idea how easy or difficult it is to get a fresh firmware copy. Especially as a lot of the time the manufacturers seem to prefer making shit way more difficult than it needs to be. I think u mentioned that the controller chip could be bad at the top of the video. Which is why I’m asking about the reuse of the same controller firmware. I really enjoy watching you work as you take care to make it as easy as possible to follow along. Which is appreciated 🙏 Cheers ✌️🖖 *an added comment here after I listened to what you said a few seconds after flashing the firmware. I hadn’t gotten that far as I stopped the video when I thought I saw that it said failed. So I backed up and the commented before you more or less confirmed what I wondered about reusing the firmware 🥴. Anyhoo
sometime clearing and writing same program can solve problem. You dont have a nand memory type in these devices but nor like memory. nor memory dont have conotroler to utilise memory banks so if data banks are corrutped or are unable to be cleared it lead to errors. Thats why clearing and refreshing banks can resolve problem but you are right that it is a common sense to put verified good program inside.
Excellent video thank you for the content, I actually learned a lot from this. You don't see many people crack out the oscilloscope these days. True master repair. What I take mostly from watching your videos is that no way on earth is north bridge repair a guy to send graphics cards to for repair I've yet to see him pull out a schematic and show true knowledge about them. 🤣 Still he has skills though.
7:55 What I heard was, "This is basically a multiface controller. It controls all of your faces.". And it was so hilarious because from time to time you crack some hilarious jokes like that. Man, the video title was so ominous but I didn't want to spoil the ending. So glad you were able to fix it in the end.
@northwestrepair it happened to me in several faulty dr mos, but these were graphic cards with shorted 12v aux rail. Maybe in cards that crash and good IOR chip is not the same. Try it.
The 4 channel version of you scope (SDS 1204X-E) has a WebServer that shows a web interface of the scope. It would be simpler than having a camera to capture the waveform.
Next time try to solder a jumperwire (maybe they connects to a few resistors) to the controller pwm pins and monitor them at crash. The falty one will make a bad pulse, others will cut out at the same time. But sometimes the internal timer is bad and you can't check it. But yeah time consuming and not quite worth it.
TLDR/TLDW: He's not as experienced troubleshooting the specific AMD controller (he mentions this himself @56:53) and he also mentions he's not a specialist with AMD controllers of this specific type. That's why he doesn't like fixing AMD cards as it can be complicated for him.
According to him, he still thinks that AMD cards (at least from more recent generations) are inherently more difficult to diagnose and repair. It's arguably not a major issue for the consumer, it basically just means that repairs outside of warranty coverage can be more expensive and time consuming. Still, I would like to see AMD make improvements in this area. Electronics should be designed to be reasonably easy to diagnose and repair, whenever possible. From the consumer's perspective, most cards will never have a major failure even if you use them heavily for five years in a row. If the cards are used properly, and if they're decently well made, the first components which are likely to stop working are the fans, and those SHOULD be fairly easy to replace. Some brands make it a lot easier than others, but you can always resort to just zip-tying some case fans to the card's heat-sync, as long as there is enough space in your case, and this can work very well as long as they are powerful enough.
@syncmonism that is his opinion, and it's definitely a valid one, but another person who's worked with more AMD cards can say the same thing with NV cards if they don't work with it as much they do. Each shop and person has specialization, and one can say one is easier based on their experience.
The chip programming also gave the same error messages on the original chip, so it didn't actually get programmed I assume. I wonder if it was actually faulty or just corrupted firmware all along... I'm also kinda nervous about the integrity of the settings readout if there was a flaky I2C connection to the chip. Or does it do some kind of error checks on the communication?
I think you are by far the best at soldering of anyone on the net. Alex says better than factory. Not sure about that, but he is excellent at correcting other sliders mistakes. This morning he looked at a connector and the whoever had placed a large blob of solder on all the connections but since this was a Nintendo Switch he might not charge over the moon charges. Kris Fix in Churmany solders chips back with non leaded. I feel the reason for a lot of failures is no lead, but a lot of his pads have the pad pulled off, thus not the solder but board failure due to flexing.
With all due respect for your enormous knowledge of the subject, I find your videos great to dream away deep in the night, as I was allowed to do earlier when painting Bob Ros. I thank you very much.
This is the issue with some manufacturers. Even with the equipment I work on, sometimes the schematics are ruined or not present and you cant find any online. Now you just gotta be a technician and troubleshoot, if this shit were easy all the time anybody could do this kind of work. Either small or large, sometimes youre gonna have to dig and get a result either way.
@@AshnSilvercorp I agree with you, a lot bigger systems and chillers Ive worked on utilize some form of DDCs. For some boards you wouldn't really know everything you're looking at unless you have past experience to guide you, even then you don't have specifics. Thankfully (in my case) with refrigeration equipment manufacturers & controls, people will usually patch us through to their engineers to talk specifics. For these companies, I doubt there's actual 3rd party "tech support" in most cases.. Id bother em anyway.
I have similar problem with my PowerColor R9 285. The difference is my green screen crash comes intermittently. I think I'm going to check if my card use this chip and check the solder joint. Thank you for this informational and educational video.
Very educational! Good job! But what if you disable the phase from controller (the one that was faulty) and the check with the oscilloscope to see which one is it?! Then swap dr-moses, then replace controller. Wouldn’t be easier?
Nice video! It kept me glued to the screen until the end which I didn't realize was 5:45 am now where I am. 😲 Very much worth it as I learned a lot about how much of a pain in the @ss repairing modern AMD GPUs are! 👍👍 Happy 4th everyone! Be safe out there!
Very interesting to watch. I started watching your videos and I kinda want to try and fix some graphic cards myself and see where it takes me. Thank you very much for your videos. At the end you said that it the controllers are expensive. I was wondering how much?
As usual very educative vid. Would you please consider to make one specific video about testing AMD's memory. I think there is many confused tecs on this subject myself included. What software best use? What are the speciffic settings? Where to look for it? Nvidia with MODS/MATS is simple but AMD is a trully headache.
So the GFX phases are arranged like this: 10 (=10) 5 (+5=10) 9 + 1 (=10) 4 + 6 (=10) 3 + 7 (=10) 8 + 2 (=10) with the bottom 4 rows doing a clockwise sweep like a lower-case omega starting in the top-right: ω EDIT: And then I reached 32:00 in the video and understood the problem.
7:07 Expensive? Well its then most likely just because its branded. The same USB to I2C, as it is written on it, can be bought for a buck or two on Amazon fex.
Just curious about whether or not you ddu your graphics drivers off your computer so you dont keep loading driver ontop of drivers . That will corrupt the drivers your using for the particular card making it lock up or black screen or just reboots your computer? Thank you for your show. You do great work!
This, to my experience seems like something made up by people with broken or failing graphics cards or hardware in general. Most of the time the hardware is the issue, not the software (drivers).
I was having black screen on my computer while I was using it regurarly or playing games. It was not happening so often. Sometimes it ran 6 hours without any issue but in the mornings it didn't turn on. When I brought it to one of the pc repair shop in my city, my PC was giving bluscreenen during the furmark test this time. We swapped my Graphic card to another PC, and it was fine. Then used new rams, new power supply with different graphic cards on my own pc again it gave the same error. We ended up the problem occurred because of windows 11. Same components were not giving an error during furmark test on win 10 but right after the test it was giving it on win 11. It was such a dumb mf annoying problem.
Hello🤝 it was good, you have a lot of knowledge about graphics cards, especially AMD. I have Rx580 Armor 8GB OC and I don't know what it's about. I tested the memory with your Tserver and it's ok, but the problem is that it crashes during games, but furmark and heven run normally. In your opinion, can I try a new bios?
Is there a way that you can configure the phase controller to lock the phases and not move them around? I think they would have a diagnostic mode that locks them in?
I just bought a 2nd hand Gigabyte 6800xt gaming OC. I just found out the die has dried. But I have no way as of now to test it because of few reasons: 1. My RAMs are just on its way (1st I bought send me 2x8gb instead of pair of 16gb. I have to wait for 1wk for the the thermal pads since no store here my area has it, just thermal paste. I just look it now, and It has a slight elevation on 1 side. What advice can u guys give on how to put paste on it since reball not an option in my area or country. TIA
@@northwestrepairI assume he means the Command Line Interface, the Linux terminal or whatever. And just to interject, my guess is just using some sort of hdmi pass through capture card so you can also capture video signal corruptions?
Fantastic video 🎉, I lways thought that PWM chip running 5+1 meaning 10 gpu cores and 2 SOC , “system on chip “or misc “Miscellaneous” as NVIDIA likes to call and 2 memory running by a separate chip , and the phase consistent of 2 drmos always on top of each others as a pair and when one fail 2nd one almost always fail , I don’t know iam only a doctor loving tech videos , thanks for explaining ❤
So doublers cause the phases to be not coherent? This a card from power color, does all vendors (such as sapphire, xfx, asus, msi etc) use similar controllers on AMD cards that are hard to diagnose and fix? Could be if you got more AMD cards you would be more well versed dealing with such controllers.
I think worst case with doublers you would have to replace two drmosfets? That is if you could not identify which is faulty... I never repaired any AMD GPU besides swapping memory chips so....
@@kikihun9726 Interesting. Good I never had to touch these. I only deal with NV cards, I worked on one Vega64 and one 390X .. IDK either AMD is not popular here or those hold better over time.
@@kikihun9726 BTW working on Vega64 I learned about Infinion, so I knew if I ever had an AMD card to repair that has a power delivery issue I would need their I2C reader and software, but that was just a physical damage repair so never dived deeper into the subject.
@@DIYRepairHour gtx980ti🤬 One mosfet died, other one seemed okay. Replaced the dead one, oh lucky me both mosfet smoked out.🤣 That time realized this rule
@@kikihun9726 That happens on NV cards I know. but personally I've seen this just a handful of times. I think this heavily depends on the drmos manufacturer and model.
So tldr if the controller is complaining, just axe it out of the board and get a new one, better go for that than deal with the headache you had diagnosing the phases
I have a similar problem with Vega 64, but when I set clocks to minimum in afterburner it works without crash. It also works well in furmark and benchmarks with Vega 56 bios and it doesn't crash in pre 2019 games. Any idea what it might be?
hello excellent video as always a great job I can ask you where I can find that program for the vrams test the gigabyte one would be very grateful if you share it with me thanks
Is there a way to check for a dead core - short? My Vega 64 has no short on VRM, but doesnt boot at all, if I use it as a secondary card, sometimes if Im lucky with the right motherboard it will show up in device manager and tell me its device ID, I feel its most likely a dead core, wish there was a way to test so I could just throw it away if so
Great video very in deph witc is exactly what i need. Is it too much to ask for the usb programer link (ebay, amazon, ali express) I have 3 rx 400 series with problematic vram controler.
Seems a bit odd to me, if I understand correctly the process you used. First, you read the firmware (or firmware parameters? config profile?) out, and then reflashed it. If that info was corrupted when you read it, wouldn't it still be corrupt when you flashed it back, even to another chip? Maybe I am confusing firmware with config data, though. Firmware being already programmed on the chip, and the parameter/config data read/flashed (ie, you weren't reflashing firmware, just the config data). Also, how come there is such a low limit on how many times you can flash those chips? I mean, I know flash has cycle limits, but THREE? That seems rather arbitrarily low; even old EEPROMs back in the day could to hundreds to thousands of write/erase cycles. Seems like it is an excuse to sell more chips. (though I guess normally they would rarely need to be reflashed at all, but still.. the tech is far better than that; why the arbitrarily low limit?). Lastly, I was curious after seeing the issue with the chip soldering problem if the problem might have been a temp-sensitive cold solder joint on the original chip? It is a controller chip, so I wouldn't expect it necessarily to warm up more from GPU load, but maybe the board warming up from other components under load caused a problem joint to fail? Thanks for the great video, though! :) Very informative.
i read the firmware just in case. I flashed it again to make sure chip is still in read/write able mode. Otherwise it would indicate a problem. These chips can only be programmed 3 times. Its how they are made. I dont know why that is but its a fact. It may be a firmware limitation rather than hardware limitation. Dont know for sure.
Where do you get that software at 1:55? It says Gigabyte graphics card test, but i've found nothing searching for anything related to that. It seems to be linux , but again found nothing.
As a Japanese I'd say he would be "Naruto Engineer". I find it always awesome that Tony takes a lot of effort, patience and makes it out✨ instead of giving it up as "not fixable".
Someone may misunderstand: He's not meaning AMD card is too bad to use. It's just more complicated and paintaking to diagnose an error and fix it, which is why the title says i dont like "fixing" AMD cards.
AMD sucks.
@@Look_What_You_Didno.
@@Look_What_You_Did Brand loyalty is the highest form of ignorance
@@Kosphy There is one kid that licks AMD boots. Then there is the OP... Displaying a strong cognitive bias. Overlooking the repair is need, then overlooking all the missing data points the competition provides. A bias so strong only a direct statement against the manufacture would be considered.... because all the other evidence does not exist.
@@Kosphy i mean it def is since there are plenty of reasons that amd does not suck
The thing why that controller fail a lot is because it's designed to be used with ir/infineon drmos/sps and not sic649 or sic632 so they can comunicate better. Also in general for me repairing AMD cards are easier than nvidia simply because most nvidia cards i saw are dead cores instead of AMD most of them are just dying transistor or shorted drmos.
It's also the same on nvidia cards btw, MPS controller are designed to be together with MPS SPS/drmos so when they fail their protection can kick in properly
This is really interesting information. You should make a video discussing this topic.
thats why nvidia cards last longer and not die until they have a heart attack
@@accurian148I have seen more MP86990 DrMos (coupled with an MPS2888A) burn a hole through a card (mostly zotac cards) than other brands of Mosfets/controllers, causing several layers to short together and making the repair almost impossible. (Good luck reconfiguring an MPS controller to run with 1 missing phase).
@@gpudoctor doesn't shock me at all tbh, Zotac use cheap components and design their cards with the same philosophy unfortunately. It wouldn't be so bad but they also overclock the cards aggressively making them ticking time bombs, especially for the higher end cards.
I respect showing both the soldering failures and even things that may seem silly like forgetting to solder the dongle back in. Incredibly informative and great walk through!
DRIVECLUB never dies 😎
I'm not fixing gpu-s or electronics but I find it interesting and entertaining. Thank you for sharing your skills and good luck!
GPU = Initialism (abbreviation) of "graphics processing unit".
Plural: Two GPUs (which will be written out: ""graphics processing units")
Easy, or?:) For reference see Wiktionary.
You know how complicate this job is. Now image who are those people who invented this and made it possible.
39:17 Your first flash with the original chip failed as well, but you didn't catch it. I suspect something in the software or hardware of your flashing setup, not the target chips themselves.
Very informative video. While it's more complicated to diagnose, I don't think it matters much for the end user. Also 52°C and 70°C hotspot? Those are pretty good temps right there.
Doubling fases is common practice, I'm not a board tech, but I immediatly spotted the fact they called it 5+1 and there's actually 14, mosfets, inductors and caps. However, like you say, the 2 SOC phases arent in the 5+1 equation. The +1 part is the memory VRM, which exists out of 2 mosfets as well. And the remaining 5 VRM phases are doubled and divided over 10 mosfets.
I don't think so, the plus meaning independent voltage regulator is loop 2 that is soc not mvdd that controller by other pwm maybe ncp81172n
There is a little devil inside those phase doublers wich sifhts the phases around every time when the repair man is trying to find the right one :)
Wow that was a nightmare... But you figured it out! You videos are very educational and entertaining, thank you!
A lot of 6000 series cards made during the pandemic, mainly the original 6700/6800/6900 variants, suffered a lot from subpar quality components being used because less people were working In the factories, and there was much less oversight to ensure quality control. I've seen many of these cards fail due to many issues including bad silicon, faulty bios chips, and just subpar quality builds Even with brands like Sapphire, PowerColor, XFX, and ASRock which supply the majority of AMD's GPUs due to these companies being strictly AMD only. I would actually suggest to anyone owning any card from the 6000 series to look for the 6x50 variant of any card because the build quality will be much better.
PowerColor and AsRock aren't much quality. They aren't bad but they aren't quality either
@@gabrielgabriel9779 PowerColor is actually one of the two top brands, especially the Red Devil series is on par with Sapphire's Toxic/Nitro+.
@@gabrielgabriel9779 asrock is the best of the companies for quality, I use asrock boards and gpus exclusively better performance and longevity.
@@gabrielgabriel9779 if AsRock isn't quality i dont know what is. Especially when it comes to boards they use high quality components where it matters as caps and mosfets while keeping a reasonable price for the customer.
@@gabrielgabriel9779 I got a ASRock Challenger Pro D RX 6600 XT when it came out and it is one of the best ones around. Used it for gaming and mining. VRAM temps never came above 70C and gpu usually stayed around 50/55C fans hardly running.
ASRock ever since they entered the gpu market have produced good solid cards.
The RX 6600 XT now runs in my sons rig without any issues. Times have changed Asus has gone down and ASRock is now with the premium brands.
I don't know if it is just reflections but the cap to the left of phase 3 you identified looks to be bulged.
It just looks a little different than the other caps, might be it is not sitting fully flat compared to the rest so reflecting the light differently.
Maybe it does look a little domed at 34:29 when the light moves across it.
I noticed the same watching the video and was wondering whether I'm wrong on that... 😊
you're right, it's bulged
@@Ren-kei It's like "Sacon FZ cap fiasco" of the GeForce 7-series days all over again! (and for ATI at that time, even when AMD bought them already in 2006, I guess Radeon X1000 series? Or very-early Radeon HD series? Such as Radeon HD 2000 series?)
uurgghh these AMD card are the bane of my existence, RX 5000 series upward are a joke to fix, even with the programmer it's still a pain
I have a RX6700XT at least I know now where to send it ;-) Great video like that you have the time and patients to go over every thing
When flashing at 39:18 it said “Device 1 User section programming failed” followed by many more errors.
Also isn’t it unwise to dump and then reuse a potentially corrupted firmware? Of course I have no idea how easy or difficult it is to get a fresh firmware copy. Especially as a lot of the time the manufacturers seem to prefer making shit way more difficult than it needs to be.
I think u mentioned that the controller chip could be bad at the top of the video. Which is why I’m asking about the reuse of the same controller firmware.
I really enjoy watching you work as you take care to make it as easy as possible to follow along. Which is appreciated 🙏
Cheers ✌️🖖
*an added comment here after I listened to what you said a few seconds after flashing the firmware. I hadn’t gotten that far as I stopped the video when I thought I saw that it said failed. So I backed up and the commented before you more or less confirmed what I wondered about reusing the firmware 🥴. Anyhoo
sometime clearing and writing same program can solve problem. You dont have a nand memory type in these devices but nor like memory. nor memory dont have conotroler to utilise memory banks so if data banks are corrutped or are unable to be cleared it lead to errors. Thats why clearing and refreshing banks can resolve problem but you are right that it is a common sense to put verified good program inside.
Excellent video thank you for the content, I actually learned a lot from this. You don't see many people crack out the oscilloscope these days. True master repair.
What I take mostly from watching your videos is that no way on earth is north bridge repair a guy to send graphics cards to for repair I've yet to see him pull out a schematic and show true knowledge about them. 🤣
Still he has skills though.
bro just use datasheet of IR35217 to find phase 3
you lose hour for thing can do in 2 min
your vids are so calming, yet educating and somehow entertaining.
7:55
What I heard was, "This is basically a multiface controller. It controls all of your faces.". And it was so hilarious because from time to time you crack some hilarious jokes like that.
Man, the video title was so ominous but I didn't want to spoil the ending. So glad you were able to fix it in the end.
😂
amazing repair, I think you can diagnose faulty dr mos by measuring the V in pin and check if it has lower resistance than the others
Interesting.
You're saying it's internal resistance on vin will lower it it's faulty ?
@northwestrepair it happened to me in several faulty dr mos, but these were graphic cards with shorted 12v aux rail. Maybe in cards that crash and good IOR chip is not the same. Try it.
Can't believe I can watch a 57 mins video of a fantastic repair job on a GPU.
Really great video, thank you. I especially like that you share all the issues and traps that most people would surely also encounter, great work!
The 4 channel version of you scope (SDS 1204X-E) has a WebServer that shows a web interface of the scope. It would be simpler than having a camera to capture the waveform.
let's hack the scope
With CS degree and a bit of electronics knowledge it was really entertaining to follow! Thanks for the amazing content! You are a humble Jedi!
Next time try to solder a jumperwire (maybe they connects to a few resistors) to the controller pwm pins and monitor them at crash. The falty one will make a bad pulse, others will cut out at the same time. But sometimes the internal timer is bad and you can't check it. But yeah time consuming and not quite worth it.
1. Your knowledge is fantastic. 2. You are also fantastic educator! now i understand why amd cards are......to fix! Regards
TLDR/TLDW: He's not as experienced troubleshooting the specific AMD controller (he mentions this himself @56:53) and he also mentions he's not a specialist with AMD controllers of this specific type. That's why he doesn't like fixing AMD cards as it can be complicated for him.
According to him, he still thinks that AMD cards (at least from more recent generations) are inherently more difficult to diagnose and repair. It's arguably not a major issue for the consumer, it basically just means that repairs outside of warranty coverage can be more expensive and time consuming. Still, I would like to see AMD make improvements in this area. Electronics should be designed to be reasonably easy to diagnose and repair, whenever possible.
From the consumer's perspective, most cards will never have a major failure even if you use them heavily for five years in a row. If the cards are used properly, and if they're decently well made, the first components which are likely to stop working are the fans, and those SHOULD be fairly easy to replace. Some brands make it a lot easier than others, but you can always resort to just zip-tying some case fans to the card's heat-sync, as long as there is enough space in your case, and this can work very well as long as they are powerful enough.
@syncmonism that is his opinion, and it's definitely a valid one, but another person who's worked with more AMD cards can say the same thing with NV cards if they don't work with it as much they do. Each shop and person has specialization, and one can say one is easier based on their experience.
The chip programming also gave the same error messages on the original chip, so it didn't actually get programmed I assume. I wonder if it was actually faulty or just corrupted firmware all along... I'm also kinda nervous about the integrity of the settings readout if there was a flaky I2C connection to the chip. Or does it do some kind of error checks on the communication?
Watching you fixing AMD cards is far more entertaining and educating than the nVidia ones.
I think you are by far the best at soldering of anyone on the net. Alex says better than factory. Not sure about that, but he is excellent at correcting other sliders mistakes. This morning he looked at a connector and the whoever had placed a large blob of solder on all the connections but since this was a Nintendo Switch he might not charge over the moon charges. Kris Fix in Churmany solders chips back with non leaded. I feel the reason for a lot of failures is no lead, but a lot of his pads have the pad pulled off, thus not the solder but board failure due to flexing.
With all due respect for your enormous knowledge of the subject, I find your videos great to dream away deep in the night, as I was allowed to do earlier when painting Bob Ros. I thank you very much.
Dang it's sad to know that stuff like this could potentially be cut in half the time just by having a board view...
Schematics or die.
This is the issue with some manufacturers. Even with the equipment I work on, sometimes the schematics are ruined or not present and you cant find any online. Now you just gotta be a technician and troubleshoot, if this shit were easy all the time anybody could do this kind of work. Either small or large, sometimes youre gonna have to dig and get a result either way.
@@Syphonfllter that would be so, but that works if you're experimenting. If it was a job, it's gotta be schematics and board view or die.
@@AshnSilvercorp I agree with you, a lot bigger systems and chillers Ive worked on utilize some form of DDCs. For some boards you wouldn't really know everything you're looking at unless you have past experience to guide you, even then you don't have specifics. Thankfully (in my case) with refrigeration equipment manufacturers & controls, people will usually patch us through to their engineers to talk specifics. For these companies, I doubt there's actual 3rd party "tech support" in most cases.. Id bother em anyway.
I have similar problem with my PowerColor R9 285. The difference is my green screen crash comes intermittently. I think I'm going to check if my card use this chip and check the solder joint. Thank you for this informational and educational video.
I have the same issue with my rx 460 too bad I throw it to the bin. I learned a lot of your videos thanks!
Very educational! Good job! But what if you disable the phase from controller (the one that was faulty) and the check with the oscilloscope to see which one is it?! Then swap dr-moses, then replace controller. Wouldn’t be easier?
or you can desolder the coil and try again the card, it can run without 1 phase
damn, that was tuf. Thanks a lot for sharing this work. Have some rest man i was getting tired just by watching you working on it, so what about you.
Nice video! It kept me glued to the screen until the end which I didn't realize was 5:45 am now where I am. 😲
Very much worth it as I learned a lot about how much of a pain in the @ss repairing modern AMD GPUs are! 👍👍
Happy 4th everyone! Be safe out there!
Awesome, you did great job, fixing this GPU.
Thanks!
I'm very glad that you want to fix and on your channel.Thanks so much
Very interesting to watch. I started watching your videos and I kinda want to try and fix some graphic cards myself and see where it takes me. Thank you very much for your videos. At the end you said that it the controllers are expensive. I was wondering how much?
Good luck!
As usual very educative vid. Would you please consider to make one specific video about testing AMD's memory. I think there is many confused tecs on this subject myself included. What software best use? What are the speciffic settings? Where to look for it? Nvidia with MODS/MATS is simple but AMD is a trully headache.
So the GFX phases are arranged like this:
10 (=10)
5 (+5=10)
9 + 1 (=10)
4 + 6 (=10)
3 + 7 (=10)
8 + 2 (=10)
with the bottom 4 rows doing a clockwise sweep like a lower-case omega starting in the top-right: ω
EDIT: And then I reached 32:00 in the video and understood the problem.
Is using doublers a cost saving measure? What would be better off?
It would be better to have second controller.
STOP RIGHT THERE CRIMINAL SCUM! YOU SHOULD HAVE PAID THE FINE HRYRUYYU YRA
happen to know if there's there a version of MODS that supports 8xxx/9xxx series cards? (think 2006/7/8)
The fact that this video gradually built into this intense and fire environment got me, I'm high on weed and this shit is lit 🔥
What a cluster F. Not you the AMD. So the controller itself. My goodness mon, you are persistent. Good job.
Great video, lots of valuable knowledge. Thank You!
I have had a RTX 3080 since November 2020 and it's still running strong.
This is hella of a cooler, one of the best with the STRIX.
it is much easier to mark on the light capacitors than on the dark and thermal pad residue chips.
watching this now because my RX6700XT is giving the same symptoms. i'm fucked.
😎Thanks for your hard work Tony you're the best!🤓
My pleasure!
From Vietnam with big love..
U are very talented and good at what u do, keep up the good work Tony !
7:07 Expensive? Well its then most likely just because its branded. The same USB to I2C, as it is written on it, can be bought for a buck or two on Amazon fex.
Just curious about whether or not you ddu your graphics drivers off your computer so you dont keep loading driver ontop of drivers . That will corrupt the drivers your using for the particular card making it lock up or black screen or just reboots your computer? Thank you for your show. You do great work!
This, to my experience seems like something made up by people with broken or failing graphics cards or hardware in general. Most of the time the hardware is the issue, not the software (drivers).
maybe you can share what tools and equipment you using for such tremendous fixing skills
After reading the title: "ok, let's see why NWR, should be a fun..."
Looks at video runtime: "50+ minutes?! Oh boy, so that's why"
I was having black screen on my computer while I was using it regurarly or playing games. It was not happening so often. Sometimes it ran 6 hours without any issue but in the mornings it didn't turn on. When I brought it to one of the pc repair shop in my city, my PC was giving bluscreenen during the furmark test this time. We swapped my Graphic card to another PC, and it was fine. Then used new rams, new power supply with different graphic cards on my own pc again it gave the same error. We ended up the problem occurred because of windows 11. Same components were not giving an error during furmark test on win 10 but right after the test it was giving it on win 11. It was such a dumb mf annoying problem.
Hello🤝 I greatly appreciate your patience and knowledge WOW👍👍👍👍👏👏👏👏👏. P.S. What is this program for checking memory chips? Regards👋👋👋👋👋👋
23:00 was the left hand side, second from top. then a later it was the first from top. I said to my self, now I gotta watch all of it.
I'm just curious... he just replaced the paste and pads ,and .. how the hell he screwed the vrm controller? Static can corrupt firmware chip?
I feel your frustration. But extremely interesting repair. Thank you!
Hello🤝 it was good, you have a lot of knowledge about graphics cards, especially AMD. I have Rx580 Armor 8GB OC and I don't know what it's about. I tested the memory with your Tserver and it's ok, but the problem is that it crashes during games, but furmark and heven run normally. In your opinion, can I try a new bios?
As an Amd fan well like them both team red and green awesome vid keep up the great work !
Thank you so much for this long videos! I love it ❤ learning so much about graphics cards 😊👍
Is there a way that you can configure the phase controller to lock the phases and not move them around? I think they would have a diagnostic mode that locks them in?
I just bought a 2nd hand Gigabyte 6800xt gaming OC. I just found out the die has dried. But I have no way as of now to test it because of few reasons: 1. My RAMs are just on its way (1st I bought send me 2x8gb instead of pair of 16gb. I have to wait for 1wk for the the thermal pads since no store here my area has it, just thermal paste. I just look it now, and It has a slight elevation on 1 side. What advice can u guys give on how to put paste on it since reball not an option in my area or country. TIA
please check 22:56 vs 23:26
cause something strange has happen
first two lanes on the right are switching places..lol..why
My 5500XT probably have the same problem. Display keeps crashing and no driver update ever fixed it.
Actually, you didn't flash the firmware before swapping the controller. It shows the same programming failures, but you didn't see them 😅
How do you screencapture so beautifully the CLI?
What's cli ?
@@northwestrepairI assume he means the Command Line Interface, the Linux terminal or whatever.
And just to interject, my guess is just using some sort of hdmi pass through capture card so you can also capture video signal corruptions?
The first firmware flash failed also... you missed the errors.
so what was the cause of the problem. User have only changed thermal interface on the card. How does it effect the controller~? Some static idk ?
the firmware became corrupted he literally tells you this in the video
Fantastic video 🎉, I lways thought that PWM chip running 5+1 meaning 10 gpu cores and 2 SOC , “system on chip “or misc “Miscellaneous” as NVIDIA likes to call and 2 memory running by a separate chip , and the phase consistent of 2 drmos always on top of each others as a pair and when one fail 2nd one almost always fail , I don’t know iam only a doctor loving tech videos , thanks for explaining ❤
Amazing work.
Thanks a lot!
So doublers cause the phases to be not coherent?
This a card from power color, does all vendors (such as sapphire, xfx, asus, msi etc) use similar controllers on AMD cards that are hard to diagnose and fix?
Could be if you got more AMD cards you would be more well versed dealing with such controllers.
I think worst case with doublers you would have to replace two drmosfets? That is if you could not identify which is faulty... I never repaired any AMD GPU besides swapping memory chips so....
If one mosfet dies, it kills the doubler and the other mosfet. No matter what, always the case.
@@kikihun9726 Interesting. Good I never had to touch these. I only deal with NV cards, I worked on one Vega64 and one 390X .. IDK either AMD is not popular here or those hold better over time.
@@kikihun9726 BTW working on Vega64 I learned about Infinion, so I knew if I ever had an AMD card to repair that has a power delivery issue I would need their I2C reader and software, but that was just a physical damage repair so never dived deeper into the subject.
@@DIYRepairHour gtx980ti🤬
One mosfet died, other one seemed okay. Replaced the dead one, oh lucky me both mosfet smoked out.🤣 That time realized this rule
@@kikihun9726 That happens on NV cards I know. but personally I've seen this just a handful of times. I think this heavily depends on the drmos manufacturer and model.
So tldr if the controller is complaining, just axe it out of the board and get a new one, better go for that than deal with the headache you had diagnosing the phases
I think i am hearing a 3d printer working in the background in 43:29.
I have a similar problem with Vega 64, but when I set clocks to minimum in afterburner it works without crash. It also works well in furmark and benchmarks with Vega 56 bios and it doesn't crash in pre 2019 games. Any idea what it might be?
these 1h videos are nice
23:24 No it was right, however when you looked the next time it had changed to the one on the top instead. xD
Love them qfn/dfn packages :)
How much was the controller? And how much do you charge for simular AMD card repair?
"AMD, quality so good even your technician will have to look it up to see what kinda card it is"... Several times...
It was the board fail not amd I'm afraid
Where can I get this Gigabyte testing software?
hello excellent video as always a great job I can ask you where I can find that program for the vrams test the gigabyte one would be very grateful if you share it with me thanks
discord'
Do u have a video where you add more v ram to a gpu ?
What tool do you use to test memory?
So you can only flash 3 times?
Is that just AMD or Nvidia too?
Is there a way to check for a dead core - short? My Vega 64 has no short on VRM, but doesnt boot at all, if I use it as a secondary card, sometimes if Im lucky with the right motherboard it will show up in device manager and tell me its device ID, I feel its most likely a dead core, wish there was a way to test so I could just throw it away if so
You cannot test core for short.
Great video very in deph witc is exactly what i need. Is it too much to ask for the usb programer link (ebay, amazon, ali express) I have 3 rx 400 series with problematic vram controler.
www.digikey.com/en/products/detail/infineon-technologies/USB005/5244633
Extremly awesome video
Thank you very much!
How often do you see PowerColor cards?
How often are those cards "modded" by the owner, before you see them?
Seems a bit odd to me, if I understand correctly the process you used.
First, you read the firmware (or firmware parameters? config profile?) out, and then reflashed it. If that info was corrupted when you read it, wouldn't it still be corrupt when you flashed it back, even to another chip? Maybe I am confusing firmware with config data, though. Firmware being already programmed on the chip, and the parameter/config data read/flashed (ie, you weren't reflashing firmware, just the config data).
Also, how come there is such a low limit on how many times you can flash those chips? I mean, I know flash has cycle limits, but THREE? That seems rather arbitrarily low; even old EEPROMs back in the day could to hundreds to thousands of write/erase cycles. Seems like it is an excuse to sell more chips. (though I guess normally they would rarely need to be reflashed at all, but still.. the tech is far better than that; why the arbitrarily low limit?).
Lastly, I was curious after seeing the issue with the chip soldering problem if the problem might have been a temp-sensitive cold solder joint on the original chip? It is a controller chip, so I wouldn't expect it necessarily to warm up more from GPU load, but maybe the board warming up from other components under load caused a problem joint to fail?
Thanks for the great video, though! :) Very informative.
i read the firmware just in case.
I flashed it again to make sure chip is still in read/write able mode. Otherwise it would indicate a problem.
These chips can only be programmed 3 times. Its how they are made. I dont know why that is but its a fact.
It may be a firmware limitation rather than hardware limitation. Dont know for sure.
the first time you tried to flash it failed also, i noticed it in the video
Where do you get that software at 1:55? It says Gigabyte graphics card test, but i've found nothing searching for anything related to that. It seems to be linux , but again found nothing.
3d printer going crazy in the background lol
3:29
I used to think it's an irish of an eye.
But since i cannot un-hear things, it's a "hairy donut" now.
Oh well
Keep up the good work!👍